Compensating for VFO drift

[MrBungle]

I've seen that used before. I think it was a trick with the LM723. You could use the pass transistor as a heater and the current sense transistor as a temperature sensor and the opamp as a comparator and use this to elevate and control the temperature of the device (or some variation of that scheme) for the sake of using the voltage reference in it as a precision reference.

Another great IC. I love that one. A Bob Widlar special. RIP.

[MrBungle]

Small update on this. Haven't had much time to do anything but did a few experiments. I've canned the environmental chamber for now to concentrate on getting it stable at the initial condition. I have however got this down to a relative fine art. Key to success are:

1. Use a type 6 material core for the L (30Hz/min) or air cored (8Hz/min). Paper core i.e. rolled A4 paper laminated with superglue is amazing. The stuff sets like concrete and it's like it's not even there.
2. Use QUALITY C0G (Vishay MLCC) capacitors for the rest of it. Polystyrene are bad. Mica are bad.
3. 2n4416 and NE602 oscillators have little difference in drift which was a surprise. 8Hz/min for 4416 vs 9.5Hz/min for the NE602. 602 is more convenient and a nice mixer so will just use that.
4. Layout has little effect on stability.
5. Varactor tuning is very stable if you don't use varactors and use 1n4001 instead. All real varactors I tried are less temperature stable than any 1n4001 I tried. Tuning range is better on the varactors though but not an issue as I only want to cover 45KHz of the 40m band with this.
6. Use quality pots (Bourns 10 turns are good and about the same price as an old tuning capacitor). Cheap carbon ones are drifty as hell.
7. Don't test anything unless it's in a box with the lid on and has had at least 20 minutes to cool down from the last soldering attack! This puzzled me for a bit until I realised.
8. Spectrum coils are total crap for VFOs. You can even hear BPFs drift as they change. Not a fan!

I'll look at absolute temperature compensation in the near future. The main point from the above is to find a design which is reproducible, uses modern readily available parts (only from RS/Farnell, not specialists), is totally 100% analogue, very stable and orders of magnitude less power hungry than a DDS.

I'll detail this receiver at some other point when I'm happy with it. Might even post it to Sprat.

Objectively I want to build a stable receiver with no ICs in it at all as the end game, then do the same with a separate transmitter.

[Al (astral highway)]

Originally Posted by MrBungle

7. Don't test anything unless it's in a box with the lid on and has had at least 20 minutes to cool down from the last soldering attack! This puzzled me for a bit until I realised.

Very interesting research there, thank you for sharing. This is a subject of great interest to me.

I wonder that (silver) mica capacitors are actually potentially very good indeed, certainly the military spec ones. Also, it's worth putting several capacitors in parallel in an LC tank circuit, to share the current between them and hence reduce the self-heating effect and the drift that arises from this.

Worth remembering that simple LC tank circuits were at the heart of broadcast transmitters going back 80 years and more, and achieved very stable carrier waves with good design.

Carbon resistors permanently shift in value after soldering, even skilfully.

I have just 'built' a precision resistor of value 1R8 and because of this trait, as well as the tiny differences in tolerance between them, I made it up from eight separate 5% tolerance resistors.

[MrBungle]

Problem I found with silver mica ones is that some of them are really good, some are terrible and you just don't know what you're getting. I have been using new Cornell Dubilier mica capacitors. Nice 1% units which I have been using for measuring inductors with. Didn't get the consistency problems with the Vishay C0G ones so veered in that direction quickly. The Vishay C0G ones are SMD ones with leads attached and dipped so are state of the art in theory. They are Vishay K series https://www.vishay.com/docs/45171/kseries.pdf

I have already distributed the main bulk capacitance across at least 2 capacitors each. That does indeed help quite a bit. Think I read that in W1FB Design Notebook. Good tip.

Broadcast transmitters were indeed incredibly stable. They had a lot more skill building them than I do

Carbon composition resistors should be universally thrown away if you ask me (apart from for non critical pulse applications). I've tried and measured different resistors in high Z bias circuits on another project to see if that made any difference. 0.25W carbon film resistors are crap so I've eliminated these from my stock now. Building up an inhumanly large stock of TE LR1F 600mW resistors which are 50ppm/oC at the moment. RS are selling some values stupidly cheap.

Did an extreme experiment - get a generic 47 ohm carbon film resistor and attach it to a DMM, then go wave it over the gas hob while in ohms mode. I got one down to 10 ohms just warming it up. LR1F caught fire before the resistance changed!

[Bazz4CQJ]

Originally Posted by MrBungle

2. Use QUALITY C0G (Vishay MLCC) capacitors for the rest of it. Polystyrene are bad. Mica are bad.

Very interesting bit of work. I wouldn't dismiss polystyrene caps entirely. In your VFO, you're managing with a varicap for tuning, but the Q will not be that great. I think that in a receiver RF amp stage, you might want to use a good air-spaced tuning cap to get a good Q, and that will have a positive temp coef and so you could use a polystyrene (neg coef) in the circuit to achieve some compensation.

My recollection is that, in the past, silver mica were noted for long-term stability, but I'm not sure where mica capacitors sit in the scheme of things today.

And... your transformer testing is progressing well ?

B

[Al (astral highway)]

Originally Posted by Bazz4CQJ

My recollection is that, in the past, silver mica were noted for long-term stability, but I'm not sure where mica capacitors sit in the scheme of things today.

I have usesd the Sangamo military spec 1% dipped silver mica ones. My experience of them is that they are very stable...

Have you tried to get hold of an 'Oxley Tempatrimmer'??
The Yaesu FT221R VHF transceiver incorporates one. I also saw a good FET based design for a VFO using one. I've never actually seen one, only a photo, but they do pop up from time to time. One sold a few days ago!

[MrBungle]

I had some problems with polystyrene caps. Firstly they require special handling when soldering. Secondly, they have a high negative temperature coefficient which caused drift for me. Thirdly, they're really expensive.

Varicap is not ideal but there's a cost versus benefit thing going on there. I would love a nice air variable capacitor but there aren't any reliable sources of decent quality ones any more for a reasonable price. Even the ones from Spectrum are now £20. On top of that, you really have to have to have a reduction drive mechanism (another £15 or so) and lots of mechanical hardware. That puts the price to around £40 for just the mechanical section of a workable VFO. This is more than the entire target expenditure on the final receiver which is around £35.

Compare to a Bourns 10 turn 10k at £13.22 from RS (1559 in stock, current line product) which requires approximately a 9.5mm hole and a knob and that's it.

The objective here is to be able to build something entirely with current line, fully available parts in theory.

I figured that even if I got marginal stability I could add a huff-puff stabiliser and lock the VFO on a stable frequency anyway.

Transformer testing hasn't started yet. Hopefully in a couple of weeks. The transfomers are lurking on my shelf at the moment reminding me that I have too many projects on and suddenly a heavy workload

[Bazz4CQJ]

It's interesting that there are two previous threads on this forum about the "desirability" of otherwise of polystyrene caps and opinions are very mixed about their reliability - probably dependent in part on how well the were soldered to start with.

Re Oxley trimmers, I think we discussed this about a year ago? I acquired a couple some years ago but haven't built a VFO since then, so have never gained hands-on experience.

I am currently building my 'Megacycle Meter' GDO clone using an Acorn valve ("DC" to 200MHz ?). This has an air-spaced tuning cap and will have coupling capacitors soldered on to the valve pins. The sort of accuracy and stability required from a GDO is such that I think it can tolerate thermal drift; at least, I've never seen a GDO circuit with any degree of temperature compensation.

B

[Radio Wrangler]

Polystyrene capacitors need to be used carefully.

Up to about 60C they exhibit a significant temperature coefficient. At about 70 to 80C they exhibit a sudden, permanent shift in capacitance. Soldering too long can shift only part of the dielectric, near the lead-out wires.

Their Q is very good and the tempco nicely compensates certain ferrites like Siemens N28 material.

Everyone knows the silver-and-transparent roll-up jobs, but there were also radial leaded ones encapsulated in plastic cases. We used to use 1% ones in precision filters... and they went through the standard wave solder line and the 'dishwasher' after it. SMT and reflow would be right out, so they are best thought of as dying availability.

Good quality ex-military air variables crop up at amateur radio rallies, but usually need a ride in an ultrasonic cleaner.

For a supreme quality capacitor and a matching drive gearbox all integrated together, take a BC221 apart.

The end of production of the tempatrimmer and its cheaper (well, less expensive) cousin, the thermatrimmer was a problem. Some decades ago I designed a thermistor bridge-and-varactor circuit to make a variable tempco capacitor. But the sneakt bit is that it was designed around making the adjustment process easy. I put it in the oscillators chapter of the ARRL handbook... it's sitting waiting to be noticed in the 1995 edition and later. This can be built for whatever capacitance and frequency range you want, and from current components.

David

[stuarth]

I read an article many years ago where the basic premise was that with semiconductor oscillators, changes in the active device characteristics with ambient temperature caused frequency drift, whereas with valve oscillators, the valve was more stable, but the heat from the valve caused drift in all the other components.

So the author's solution was to use a very low power valve (a hearing aid valve?) for a low drift oscillator. Russian miniature "rod" valves are available, cheap, and (some) are very low power, for example, the 1j24b filament takes only 13mA at 1.2V, within the realms of battery power.

Stuart

[Radio Wrangler]

It's a theory. but back at the time there had been a lot of work done on the frequency stability of VFOs and valves were the only show in town. When the 2N3819 appeared it was a VERY poorly controlled device in terms of unit to unit variability and temperature caused variability. The rate at which people switched over to them was dramatic. That included amateurs and even those professionals who could afford massive construction and 'cold ovens' and oxley compensators.

It looks like the heat effects were much worse than the parametric variation effects. You can design to keep heat out of your tank. You can also design to keep transistor or valve parametric variation out of your tank.

David

[MrBungle]

Yes. I'm keeping standing current, voltage low and coupling light. That appears to help considerably with drift. I've seen some discrete VFO designs that pull 50-60mA. I'm down to 2.5mA with the NE602 and most of that bias for the mixer inside it.

Killers are the really high Ids JFETs. Those can put out a few tens of mW of heat. Modern 2N3819 seems ok, as does 2N4416. But the difference is marginal compared to a BJT inside the NE602.

That valve still takes 16mW of power with no buffering which is going somewhere.

[Radio Wrangler]

There are two extremes to the spread of VFO designs, those designed for low phase noise (which take thunderous amounts of power and drift like the devil is after them) and those designed for low drift (which can be somewhat noisier than their power hungry cousins)

David

[MrBungle]

Yes I'd sort of come to roughly the same conclusion that there were numerous horrible tradeoffs here. Part of the fun though. I have no ability to measure phase noise (yet) and drift is more important to me at the moment as I'm trying to pull out CW signals and this will be heading through a narrow filter. I don't want to try an ant on a moving train as my CW sucks enough already

Did some more work and have found some more interesting compensation tricks:

1. Don't use a SA612. Each one is different. Fine for XO/VXO though.
2. Material 7 toroids seem to compensate well for air variable capacitors
3. Forget trying to compensate varactors - drift depends on voltage (grr). Some literature in EMRFD around compensating for this.
4. Material 6 toroids seem to compensate well for polyvaricon capacitors
5. Put all my Spectrum coils away in the cupboard where I can't try and be lazy and use them. They're rubbish for VFOs. Fine for double tuned circuits etc though.

However, I'm now going to throw this all under a bus for a few weeks and look at VXOs particularly using resonators. I had some luck with a "super VXO" in a test receiver I built. That was solid as a rock but the tuning range was a bit low (6997-7020KHz with a pair of 7030 crystals)

[G0HZU_JMR]

You can often predict phase noise performance quite well using Leeson's equation. In the past I designed a simple 'trainer' oscillator to show people some basic oscillator theory. This 10.7MHz oscillator used an LC resonator and a cheap 50R MMIC gain block.

It's possible to guess the phase noise just from the NF of the MMIC, the power it puts into the resonator, the loaded Q (and loss) of the resonator can be simulated on something like RFSIM99 and the flicker corner frequency can be guessed. This can all be added to Leeson's equation to predict the phase noise at carrier offsets of 1kHz through to maybe 200kHz. The exact frequency of oscillation can be predicted with a quick VNA measurement or you could just simulate it to get a fairly accurate estimate.

Even a modest design like this can produce low phase noise at offsets of 25kHz to 200kHz that would totally outclass something like a classic HP8640B sig gen. i.e. an oscillator like this could be used for critical receiver tests at offsets of 100kHz. I've still got it somewhere, it wasn't designed for low drift but I suspect it will be a few Hz/minute after a warmup. But it won't be stable if the ambient temperature changes a few degrees. From memory it achieved -173dBc/Hz at 100kHz offset and this is within a few dB of theory. Probably 15-20dB better than an HP8640B at this offset

[G6Tanuki]

Interesting that you mention VXOs and the "Super VXO" - I must get round to playing with one of them someday.

http://electronics-diy.com/electroni...tic.php?id=930 for those who don't know what we're talking about.

[dalekmoore2007]

I don't know if you know of this IC there are others in the range higher frequency and which i think are newer versions but the TLC1799 go's from 1khz up to 30 MHZ
https://www.youtube.com/watch?time_c...&v=PNpb3vWbGRc
I am trying it on my monitor for variable line rate but same sort of idea apart from its a square wave and for what it does very simple circuit .
very tiny pain to solder but if i can do it with my old eyes any one can.

[G8HQP Dave]

I found this. Useful introduction to oscillator noise. I remember reading Leeson's book on oscillators about 15 years ago and found that it opened my eyes to things I had never really understood before.

[Terry_VK5TM]

LTC1799's are as drifty as anything, temperature sensitive, voltage sensitive and of no use for a stable VFO.

Been there, tried that and gave it up as a bad joke.

[G6Tanuki]

VFO design seems to truly be a mix of science and voodoo: sometimes the Obvious and the Scientific is trumped by the Crazy and the Bizarrely Cunning!

Example: the WWII-era Collins TCS-8 receiver uses a 12A6 beam-power output-valve as its local-oscillator, along with a massively-widely-spaced variable capacitor. I've always thought that minimising power-dissipation throughout the LO stage was important for stability, but clearly this is not always the case! I'd love to have been able to get inside the heads of the designers to find out the reasons for their choice.